You don't want to use BWA, the results will be crap. Use bismark or bison (the former is easier to setup and more flexible, the latter faster and more accurate).
You don't want to use BWA on its own, as you want to align your bisulfite treated reads against an in-silico bisulfite treated genome. You will also want a script that summarizes methylation from the alignments. Aside from bismark and bison you should look at the neat BWA-meth wrapper that brentp recently published on Arxiv.
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Bismark also has Perl scripts that allow you to calculate percentage methylation values (beyond just doing the alignment).
Presumably, you will also want to find differentially methylated regions. I would recommend using methylKit for this, but you can also use COHCAP since it looks like you used a targeted BS-Seq library:
http://sourceforge.net/projects/cohcap/
http://www.ncbi.nlm.nih.gov/pubmed/23598999
BTW, COHCAP also has a Bioconductor version (but it is a devel release). Feel free to try that if it helps:
http://bioconductor.org/packages/devel/bioc/html/COHCAP.html